Art Morozov

What it is

A deterministic rules engine for Brass Birmingham with a simple web API. It validates moves, manages sessions, and supports a lightweight client for debugging multiplayer flows.

Key achievements

• Modeled full game state with deterministic move validation and rollback support.
• Built an Axum API for sessions, moves, and player state.
• Created a lightweight web client for multiplayer flow and UI debugging.
• Added fixtures to cover tricky edge cases in tile placement and rules.

Tech

Rust, Axum, Pytorch, Python, Svelte

What it is

An automated trading bot for Polymarket. It runs rule-based strategies, manages orders safely, and includes risk controls to avoid runaway exposure.

Key achievements

• Built a strategy engine with configurable rules for market selection and execution.
• Implemented risk controls for position limits, exposure caps, and unwinds.
• Added order management with cancel and replace flows and retry-safe submissions.
• Built backtesting scripts against historical snapshots to validate edge cases and optimize trading strategies.

Tech

Rust, Tokio, Mio event loop, DPDK

What it is

A low-latency Solana trading bot built to buy newly launched memecoins in the same block as token creation. It consumes Solana's Shredstream directly, executes an optimized Rust hot path, and submits transactions from Frankfurt-based infrastructure to minimize network distance to validators.

Key achievements

• Built a Shredstream-driven Rust execution path that reacted to new token launches within milliseconds of the creator transaction.
• Gated execution to a trusted validator allowlist and only fired when an approved validator produced the block, improving same-block inclusion probability.
• Deployed latency-sensitive infrastructure in Frankfurt close to Solana validators to reduce network and propagation delay.
• Optimized transaction construction by minimizing instruction count and reducing serialization overhead on the hot path.

Tech

Rust, Solana, Shredstream, Low-Latency Systems, Trading Infrastructure

What it is

An AI learning platform for legal students. I owned core platform work across authentication, reliability, and product analytics to keep onboarding smooth and the system observable in production.

Key achievements

• Migrated from Azure AD B2C to a faster custom auth service with safe user migration.
• Improved session handling and login performance without breaking existing access.
• Added observability and reliability signals for auth and onboarding flows.
• Built a referral system and analytics dashboards to measure onboarding and failures.

Tech

TypeScript, Python, PostgreSQL, Azure, Auth, Observability, CosmosDB

What it is

A Rust client library for NDAX that wraps REST and WebSocket APIs with a clean typed interface, making it easy to build stable downstream trading and monitoring systems.

Key achievements

• Implemented async REST client with request signing, retries, and typed responses.
• Built WebSocket streaming with reconnect and subscription management.
• Normalized order book and trade events into consistent data models.
• Added integration tests using recorded fixtures to catch API changes.

Tech

Rust, REST, WebSockets, Async

What it is

A web app that rewrites AI-generated text into a more natural human-like tone. It includes subscription billing, persistence, and usage tracking for a real product workflow.

Key achievements

• Built a Next.js app that rewrites and refines AI-generated content with a focus on natural language and tone.
• Integrated <strong>Stripe</strong> billing and <strong>PostgreSQL</strong> persistence for subscriptions.
• Implemented processing flows with clear user progress states and usage tracking to monitor credits and output quality.
• Built a backend service to detect AI-generated content using ML scoring and heuristic signals

Tech

Next.js, TypeScript, Stripe, PostgreSQL, Langchain

What it is

A computer vision poker assistant that recognizes Texas Hold'em hands from images or video. It includes model training, inference optimization, and a desktop workflow for review and export.

Key achievements

• Built a desktop app that classifies poker hands from images and video.
• Trained a custom YOLO model and optimized inference for real-time use.
• Built tooling for data ingest, labeling, evaluation, and iteration.
• Designed a UI workflow for batch uploads, review, and exportable outputs.

Tech

OpenCV, YOLO, Python, Machine Learning

What it is

A book recommendation platform that trains and serves personalized recommendations end-to-end, from data prep and model training to API delivery and user-facing explanations.

Key achievements

• Built a Django REST API and React frontend for personalized recommendations.
• Trained a PySpark ALS model on 6M+ data points for recommendation quality.
• Shipped a full pipeline for data prep, training, and serving.
• Added explainability summaries to show why each recommendation was suggested.

Tech

Python, Django, React, PySpark, ML Pipeline